cloud stringclasses 4
values | region_id stringlengths 3 3 | city stringlengths 5 11 | deploys int64 5 115 | median_seconds_to_running int64 20 85 |
|---|---|---|---|---|
hetzner | fra | Frankfurt | 72 | 20 |
hetzner | hel | Helsinki | 63 | 20 |
hetzner | nbg | Nuremberg | 66 | 20 |
digitalocean | ewr | New Jersey | 6 | 29 |
digitalocean | lhr | London | 49 | 34 |
digitalocean | sgp | Singapore | 38 | 35 |
digitalocean | blr | Bangalore | 27 | 38 |
digitalocean | ams | Amsterdam | 52 | 40 |
digitalocean | fra | Frankfurt | 34 | 45 |
vultr | sao | São Paulo | 25 | 34 |
vultr | tlv | Tel Aviv | 33 | 43 |
vultr | cdg | Paris | 5 | 43 |
vultr | dfw | Dallas | 14 | 43 |
vultr | fra | Frankfurt | 20 | 43 |
vultr | ams | Amsterdam | 21 | 46 |
vultr | ord | Chicago | 6 | 48 |
vultr | nrt | Tokyo | 5 | 53 |
vultr | ewr | New Jersey | 17 | 54 |
vultr | icn | Seoul | 22 | 56 |
vultr | mex | Mexico City | 17 | 82 |
linode | ord | Chicago | 5 | 55 |
linode | sto | Stockholm | 11 | 65 |
linode | fra | Frankfurt | 26 | 75 |
linode | dfw | Dallas | 7 | 75 |
linode | nrt | Tokyo | 13 | 75 |
linode | ams | Amsterdam | 9 | 75 |
linode | ewr | New Jersey | 11 | 75 |
linode | sao | São Paulo | 10 | 80 |
linode | lhr | London | 115 | 85 |
linode | sgp | Singapore | 12 | 85 |
Cloud VPS Measurements
Four small measurement sets from NoctHost, a service that rents servers on Vultr, DigitalOcean, Hetzner and Linode. Every number was measured on fresh servers in November 2026.
| File | What it measures | Date |
|---|---|---|
exchange-latency.csv |
Median warm HTTPS round trip (ms) from servers in 21 cities to the public APIs of Binance, Bybit, OKX, Coinbase, Kraken and Hyperliquid. Empty cells mean the API does not serve that location. | 12 Nov 2026 |
deploy-time-by-cloud.csv, deploy-time-by-city.csv |
About 880 real server deploys: median seconds until the cloud reports the server running and until SSH accepts a login, per cloud and per city. | 26 Oct to 15 Nov 2026 |
vps-benchmark.csv |
sysbench CPU and memory, fio random and sequential disk results for comparable 2 vCPU / 4 GB plans. | 15 Nov 2026 |
gpu-llm-speed.csv |
Ollama generation speed (tokens per second) for Llama 3.1 8B and Qwen 2.5 14B on RTX 4000 Ada and L4 cloud GPUs, with hourly prices. | 15 Nov 2026 |
Method
Method, charts and updates are on the source pages:
- https://nocthost.com/exchange-latency
- https://nocthost.com/deploy-time
- https://nocthost.com/vps-benchmark
- https://nocthost.com/gpu-llm-speed
Archived copies with DOIs are on Zenodo (exchange latency: 10.5281/zenodo.24247822, deploy time: 10.5281/zenodo.24249955, LLM speed: 10.5281/zenodo.24251441).
License
CC BY 4.0. Please credit NoctHost (nocthost.com) when you reuse the data.
- Downloads last month
- 32